There are a few odd things about NewR. We should consider fixing them.
Because x0 is a ref param, the first call into Inner sets it to the root value on the first iteration, and all subsequent iterations of that call and all iterations of subsequent calls immediately confirm that the value is the root and exit without doing work. So we're mainly timing how long it takes to evaluate FF.
If one sets x0 = 1.0 before calling Inner then it does the same amount of work per call. If you set it to 1.0 inside then the iterations inside Inner also do similar work.
There's also something off with the iflag setting. One would expect one flag value to indicate proper convergence and another to indicate numerical breakdown, but there's only the one value.
NewR also fails to do output validation.
category:correctness
theme:testing
skill-level:intermediate
cost:small
impact:small
There are a few odd things about NewR. We should consider fixing them.
Because
x0is a ref param, the first call intoInnersets it to the root value on the first iteration, and all subsequent iterations of that call and all iterations of subsequent calls immediately confirm that the value is the root and exit without doing work. So we're mainly timing how long it takes to evaluateFF.If one sets
x0 = 1.0before callingInnerthen it does the same amount of work per call. If you set it to 1.0 inside then the iterations insideInneralso do similar work.There's also something off with the
iflagsetting. One would expect one flag value to indicate proper convergence and another to indicate numerical breakdown, but there's only the one value.NewR also fails to do output validation.
category:correctness
theme:testing
skill-level:intermediate
cost:small
impact:small